Abstract
In this work, we provide a Genetic-based algorithm that is used to quickly find a placement for a set of objects within a given layout such that access to these objects is optimized. The given layout describes the free locations of the objects and the object handles and the access is done through a corpus of object requests. The proposed algorithm optimizes the placement of the objects by searching through a small fraction of the search space. As a case study, we use the algorithm to find a better placement for the keyboard characters than QWERTY and Dvorak Simplified characters placements. The algorithm finds a placement that is better than both QWERTY and Dvorak Simplified by 32.68% and 15.79% respectively on the training set, and 32.71% and 15.84% respectively on the testing set. This result is achieved after searching through only 500K possible solutions, which is about 1.23 × 10-19percent only of the total search space. Both training and testing sets are extracted randomly from TED2013 v1.1 English corpus. Moreover, we release the dataset, code and experimental results on our GitHub repository.
Ali Fadel
Machine Learning Engineer II
Software engineer interested in problem solving and machine learning based solutions, likes to create content and teach others.